The article presents a mathematical model for calculation of nonstationary hydraulic and separation processes in gas centrifuge (GC) cascade multicomponent isotope mixtures. has been applied to calculate the parameters GC krypton, germanium tungsten isotopes. As result, specifics excess holdup distribution along stages identified, variations concentrations process have revealed. data obtained show that proposed is able adequately describe cascades mixtures.Highlights:Mathematical mixture...

The article substantiates the relevance of conducted research and determined scope application developed algorithm. Private problem identification tubular products slabs at Vyksa Steel Works is considered. shows analysis modern developments in field localization markings industrial products. It description algorithm symbolic images bar-code labelling on Shows lack versatility existing methods markings. A new method proposed that would combine results several different using probability...

The article shows analysis of modern developments in the field localization and identification markings industrial products. It description algorithm symbolic images bar-code labelling on This also focuses mathematical provides results introduction this at Vyksa Steel Works enterprise.

The article presents the project of developing automatic identification labeling system based on two-step method localization and recognition to track movement workpieces between warehouse production facilities JSC Vyksa Steel Works. developed application provides control rolled products with remote detection at a distance 3-20 m.

The paper considers an algorithm for localization of railway wheel markings based on preliminary symbols. parameters the operation were chosen. Studies developed have been carried out when changing angles shooting and lighting.

The paper considers an algorithm for localizing the markings of railway wheels. parameters required operation have been selected. Research has carried out to limit shooting conditions. functional model system is described. An example using developed software given.

As the title implies article describes actuality of algorithm development positioning industrial wares in-plant based on radio frequency grid for construction products tracking systems. Requirements international standards regulating processes traceability and identification are analysed. The offers a system hardware solution as well an determining current storage area. Experimental studies developed were conducted.

The main current directions of research in nanoscale images processing a high degree detail are shown the paper. A system algorithms for solving wide range tasks structural analysis industrial materials has been developed. experimental results developed have presented article. image shown.

The article shows lack of versatility in existing methods localization symbolic markings. A new method is proposed that would combine the results several different using probability theory and fuzzy logic to improve accuracy results. coefficient for algorithm introduced. It regulates degree influence going have on resulting outcome. presents experimental study uses two algorithms an image with industrial marking.

The external environment of the world is dynamically changing, which requires ability continuous learning and memorization from intelligent systems. Iterative methods neural networks have smooth convergence, so they do not allow achieving such an effect. An exception Extreme Learning Machine (ELM), allows for incremental learning. In ELM, input weights are randomly generated. between hidden output layers computed in one step, resulting a faster rate compared to traditional SLFN algorithms....

The article presents results of research filling gas centrifuge cascade by process fed into different stages. modeling was done for nickel isotope separation. Analysis the shows that concentrations light and heavy fraction flows after depend on feed stage number.

This article investigates the influence of various parameters on Alternate Incremental Learning Algorithm within context neural networks. The primary focus is three prominent factors: number neurons in hidden layer, choice activation function, and distribution range weight coefficients. An experimental approach was adopted where layer varied between 5 50, thereby providing insights into how network complexity affects learning capacity. considered functions encompass sigmoid, rectified linear...

The paper solves the problem of classifying marking symbols on a metal surface using neural networks that incrementally relearn in process operation. A set images railway wheel is used as source data. An incremental method training various convolutional investigated to identify most successful models with least forgetting old knowledge. description experiments assess accuracy given. model was selected demonstrates highest when additional method.

The article considers the problem of features selection for human activity recognition in telerehabilitation. notion telerehabilitation is considered. types physical that are necessary to conduct a indicated. A method carrying it out using wearable devices proposed. It proposed use smartphone, which includes an accelerometer and gyroscope determine performed by patient. literature review approaches given. An experiment select such classification would allow high accuracy with minimal...

Preliminary mathematical analysis of miscellaneous production data is presented in the paper. Research shows how strong impact each parameter on information picture enterprise. The article describes critical deviations characteristics from normal values can be timely identified. An algorithm based least squares method, which allows to increase efficiency and accuracy identifying problematic situations developed course research. Experimental research carried out by four metrics.

Extreme Learning Machine is a single-hidden-layer feed-forward neural network that has the advantage of high learning speed and ease implementation. The efficiency an extreme machine in human activity recognition largely depends on three parameters, such as weight matrix, hidden layer neurons number, activation functions. This study aimed at building artificial to solve problem analyze influence accuracy parameters networks input weights, functions, number layer. For experiment, open dataset...